Causes with material continuity

Recent philosophical work on causation has focused on distinctions across types of causal relationships. This paper argues for another distinction that has yet to receive attention in this work. This distinction has to do with whether causal relationships have “material continuity,” which refers to the reliable movement of material from cause to effect. This paper provides an analysis of material continuity and argues that causal relationships with this feature (1) are associated with a unique explanatory perspective, (2) are studied with distinct causal investigative methods, and (3) provide different types of causal control over their effects.


Introduction
A significant amount of the philosophical literature on causation has focused on distinguishing causal from non-causal relationships. This has generated interest in various definitions of causation, including those that appeal to statistical relations, counterfactual relationships, and mark transmission, to name a few. Despite this longstanding focus, recent work in this area has taken a different aim. Instead of distinguishing causal from non-causal relationships, this work aims to clarify differences across types of causal relationships. This captures "distinctions among causation" in the sense of clarifying how legitimate causal relationships differ from each other (Woodward 2010). The main distinctions discussed in this work include stability, specificity, and proportionality, with the point being that genuine causal relationships can exhibit these characteristics to a greater or lesser degree (Woodward 2010;Blanchard et al. 2018).
This paper argues for a causal distinction that has yet to receive attention in the literature. This distinction has to do with whether causal relationships have material continuity, which refers to the reliable movement of material from cause In this sense, to say that X is a cause of Y simply means that there are some values of X such that intervening on X and changing its values (in some specified set of background conditions) produces changes in the values of Y. This account is counterfactual in the sense that determining whether X is a cause of Y depends on counterfactuals about what would have occurred to Y under different interventions on X. Thus, a cause "makes a difference to" its effect and an effect "counterfactually depends" on its cause (Woodward 2003).
A first distinction among types of causation concerns the stability of causal relationships (Woodward 2010). Stability refers to the degree to which a generalization between X and Y would break down if the background conditions were changed in various ways (Woodward 2010). For example, the causal relationship between a gene and a disease phenotype is stable when the gene causes the disease in a wide variety of "background" conditions. An example of this is the causal relationship between the huntingtin mutation and Huntington's disease. As Kendler states "[i] f you have one copy of the pathogenic gene for Huntington's disease, it does not matter what your diet is, whether your parents were loving or harsh, or if your peer group in adolescence were boy scouts or petty criminals. If you have the mutated gene and you live long enough, you will develop the disease" (Kendler 2005, 394). Changes in other genetic, environmental, or social factors will not influence or alter this cause-effect relationship-it remains stable across these changes (Kendler 2005, 397). Compare this with the claim that a car crash (Y) was caused by the birth of the driver's paternal grandmother (X) (Lewis 1986, 216). Although X is in the "causal history" of Y, we view this causal claim as questionable, because we can imagine many situations and (different likely background conditions) in which X takes place, without reliably leading to Y. For example, this relationship breaks down in situations where X does not have children, Y is a better driver, or safer road conditions exist. Both cases reveal how stability is assessed by considering "relevant" background conditions, where these are specified in a context-dependent manner. Unsurprisingly, causal relationships that are more stable are viewed as more paradigmatically causal, more useful, and more supportive of deeper explanations. 1 A second causal distinction involves specificity, which comes in two varieties-value specificity (specificity 1 ) and variable specificity (specificity 2 ). Value specificity concerns the number of values of the effect that the cause can produce. In this case, a cause is highly specific when it provides fine-grained control over many values of its effect. An example of this is the fine grained-control that a radio dial provides over the speaker's volume and that a dimmer switch provides over the light emitted from a light bulb. This form of specificity is often associated with the explanatory importance of DNA and the "information" it contains (Waters 2007;Weber 2017;Stegmann 2014). In fact, this form of causal influence is said to explain why genes are "privileged" causal factors in biology, namely, because they provide more specific, fine-grained control than other causes. A cause that is less-specific 52 Page 4 of 17 provides coarse-grained, binary control over its effect, such as a standard light switch controlling whether a light is "on" or "off".
Variable specificity, on the other hand, has to do with the number of causes and effects in a causal relationship. The paradigm of variable specificity is a one-causeone-effect relationship-in this case, the cause produces a single, specific effect and the effect has a single, specific cause. An example of this is the "monocausal model" in medicine, in which a single gene variant, bacterium, or vitamin deficiency produces a single, unique disease (as is the case with Huntington's disease, tuberculosis, scurvy, etc.). Non-specificity of effect refers to situations in which a particular cause has many different effects such as a single earthquake causing many disastrous outcomes (water pipes to break, bridges to crumble, and fires to erupt) and a single gene producing a myriad of different effects (as in cases of pleiotrophy). Non-specificity of cause refers to situations in which different causes can all produce the exact same effect. An example of this emerged in research on Parkinson's disease when it was discovered that completely different causes were all individually sufficient to produce this same disease. 2 Non-specificity of cause captures an important form of causal complexity that is found in many sciences (Ross 2020a, b).
A third distinction among types of causation concerns proportionality, which has to do with the level of description of cause and effect. This distinction is well-captured with Yablo's pigeon-pecking example (Yablo 1992). In this example a pigeon is trained to peck at red targets. One day, this pigeon is presented with a scarlet target and this causes it to peck. Consider two different ways of describing this causal relationship: (i) Did the presentation of a red target cause the pigeon to peck? or (ii) Did the presentation of a scarlet target cause the pigeon to peck? Description (ii) is less correct because it describes the cause in a way that is less proportional to the effect-it suggests that the scarlet shade caused the pigeon to peck, when any shade of red would have caused this. The sense in which the cause in (ii) is less proportional relates to its inclusion of irrelevant detail (namely, the particular shade of red). Thus, (i) describes a more proportional relationship-red better captures the proper contrast because if a non-red color had been presented, the pigeon would not have pecked. While this case considers whether the cause is proportional to some effect, other examples consider whether the effect is proportional to the cause (Woodward 2010).
These distinctions are claimed to capture causal relationships that are more "useful" and that figure in "better" explanations (Woodward 2010). While these features do seem to be valued in various causal reasoning situations, there are interesting exceptions to this. Unlike some claims (Waters 2007;Lynch et al. 2019), causes with value specificity do not always provide better explanations. Reasons for this include the fact that coarse-grained control can be preferred and that fine-grained control is impossible if the effect variable is strictly binary. 3 Similar points can be made for variable specificity. 4 Nevertheless, the claim that these causes support higher quality explanations is a significant motivation for studying and clarifying these distinctions.

Causes with material continuity: a new distinction
Another distinction among types of causation, which has yet to receive attention in this literature, has to do with whether causal relationships have material continuity or not. Material continuity refers to a situation in which, the change in X that produces a change in Y is accompanied by the reliable transfer of material from X to Y. Examples of this include a ball rolling down an incline, blood traveling along a blood vessel, subatomic particles moving along a decay series, a carbon skeleton moving along a metabolic pathway, and an organism progressing through the steps of a developmental pathway. In some of these cases, the presence or absence of an object at a particular location (L 1 ) is causally relevant to its later presence or absence at another location (L 2 ). In order for the ball to arrive at the bottom of an incline, it must first be located at the top. In other cases, it is not the location of the object, but the object's prior state (S 1 ) that is causally relevant to a later state (S 2 ).
As a more detailed example, consider the metamorphosis of a butterfly. The stages of this developmental process include the egg, larvae, pupa, and adult. While there are a number of factors involved in the transformation of the egg into an adult butterfly one important cause of this is the presence of the initial egg itself. The egg's presence (in the right environment) is a cause of its transformation into the most immediately downstream step and so on. This causal sequence involves counterfactual dependencies between the different states that the organism is in. The egg's presence (or presence of the organism in the "egg" state) is a cause of its transformation into a larvae, which is a cause of its transformation into a pupa, which is a cause of its transformation into an adult butterfly. Relatedly, controlling whether the initial egg is present or not, controls whether a downstream butterfly is produced or not-if the organism gets held up at any step, complete metamorphosis will not occur. This sequence of causal steps involves the flow of material that is associated with the organism-particular molecules, cells, and tissues, for example-from one step to the next. 5 This is similar to metabolic pathways and factory assembly lines, in which some upstream substrate is converted into a series of downstream products (Ross 2018(Ross , 2021a. The substrate at each upstream step is causally relevant to the creation of the most immediately downstream product and there is material that reliably flows along this causal process. In other words, the same material that makes up the cause, becomes part of the downstream effect. While material such as atoms, molecules, and cells move along the steps of these causal processes, notice how this is not the case for clock-mechanism type causal systems. As will be discussed more soon, these mechanisms involve gears that causally interact, but without transmitting material (for example, atoms) through these interactions.
In clarifying what is meant by material continuity a few additional points should be mentioned. First, causes with material continuity often involve some significant amount of material carryover from cause to effect-it does not count to simply have microscopic particles or traces of dust moving from one to the other. 6 For example, trace amounts of dust that move from one mechanism gear to the next, do not count. This relates to a second point, which is that material continuity requires that the material "reliably" moves from cause to effect without being a coincidence or matter of sheer luck. The movement of material is, in a sense, "built-in" to the causal process. Third, the flow of material is largely constrained by and limited to the causal process. A large majority of the material moves from cause to effect, without moving to other external properties that are non-effects.
By capturing material continuity within an interventionist framework, this analysis differs from traditional mark transmission accounts of causation (Salmon 1984). On these latter accounts causation is defined by the transmission of a mark from cause to effect. 7 Alternatively, on the interventionist analysis I propose, the causal processes in question are not causal because they involve the flow of material or a mark. They qualify as causal because they meet the interventionist criterion and they have the additional feature of involving material flow from cause to effect, which is not required for causation. This allows us to capture the importance of material continuity in causal processes, without erroneously viewing this feature as required for causation. Viewing material (or mark) transmission as an additional, accessory feature of causation-a feature that some causal relationships have and others lack-allows the interventionist framework to accommodate a broader range of cases that we view as legitimately causal. For example, it allows us to capture the fact that some causal processes lack the flow of marks or material.
In examining this further notice that while some causal processes involve the flow of material, others do not. A first group of causal relationships that lack material continuity are those in which cause and effect (i) lack physical connection. Examples of this (i), include "absence causation" and "double prevention," in which the absence of an entity produces some effect and preventing an action allows for an effect, respectively (Beebee 2004;Schaffer 2016). An example of absence causation is when my houseplants die because I fail to water them. I have caused their death without physically interacting with them and, thus, without transmitting any material to them. In these cases, as cause and effect do not come into physical contact, there will be no transmission of material from one to the other.
A second group of examples show that causal relationships can lack material continuity even when (ii) cause and effect physically interact. Consider a sequence of dominos that fall over in series. We view this process as genuinely causal although there is no material that starts with the first domino and reliably moves to the final domino as the causal process unfolds. The plastic material of the first domino does not move into the second domino, and end up in the final "fallen" domino. In this case, the causes and effects are physically interacting, but these interactions do not reliably transmit material. Similar examples of this are colliding billiard balls, moving gears in a watch, and biological mechanisms (such as enzyme complexes) with physically interacting parts. A related set of examples are the cases Woodward (2010) uses to illustrate his distinctions among causation. Notice that there is no material that reliably moves from the huntington gene to the disease phenotype, from the radio dial to the speaker's volume, or from the presented target to the pigeon's pecking behavior. These cases involve the flow of causal influence and they may involve the flow of (what we consider) "information", but they lack the flow of material-atoms, molecules, and microstructure-from cause to effect, which is distinct from causal influence. 8 Why does it matter whether causes have material continuity or not? Causes with material continuity are not important because they provide causal explanations that are deeper, better, or more paradigmatic than causal relationships that lack this feature. Instead, these causes are important because they capture how causal relationships and causal systems in the world are different. These differences motivate distinct explanatory questions, they are associated with unique explanatory perspectives, they require study with distinct scientific methods, and they provide different types of control over the world. Appreciating this diversity is necessary for understanding the varied causal structure of the world and how we successfully study it. In the rest of this paper, I support these claims by clarifying how causes with material continuity (1) are associated with a unique explanatory perspective, (2) are studied with distinct causal investigative methods, and (3) provide different types of causal control over their effect. An analysis of these points will show the importance of appreciating the diverse causal structure of the world and how this can be captured with an interventionist framework.

Explanatory perspective: object-oriented causation
Causes with material continuity are associated with a unique explanatory perspective. Causes with this feature are likely to surface when we are focused on "objectoriented causation," in which we are interested in a particular object and how it changes over time. As the objects in these cases are typically associated with material, we view this material as "moving" through a sequence of causal steps. This is similar to what Russell refers to as the "persistence of something" through a causal relationship, such as "a person, table, a photon, or what not" (Russell 1948, 404). In these cases, there is a "constancy of quality" and "constancy of structure" through the steps of the causal process, where this can take the form of constancy of physical material. This is seen in how we track an object as it develops, matures, or transforms over time, or in how we track an object as it moves through a system. We can track an organism through developmental stages, a product through an assembly line, blood through the vessels of an organism, and water through the pipes of a city. In contrast, notice how these examples of "object-oriented" causation differ from mere "causal influence," in which an effect ricochets through the world without involving the reliable flow of material. For example, the spread of a fascinating news story through a community is clearly a causal process, although it need not involve the movement of material or any "object" through these steps. Similarly, Woodward's causal distinction examples also lack this object-oriented perspective (Woodward 2010). We do not identify any object moving from the huntington gene to the disease phenotype, from the radio dial to the volume output, or from the red target to the pigeon's pecking behavior.
This object-oriented perspective is further illustrated with two main categories of causes with material continuity: (i) manufacture and (ii) movement cases. 9 Manufacture cases involve changes in an object's form or constitution, while movement examples involve changes in an object's spatial location. Manufacture cases involve tracking some starting material as it undergoes a sequence of transformations, which culminate in a final product. These are associated with processes such as metamorphosis, synthesis, development, growth, transformation, maturation, and so on. Movement cases, on the other hand, involve tracking an object's change in location, such as movement of blood through vascular pathways, water flowing along channels, and a ball rolling along an incline. These cases are associated with "object-oriented" explanatory why-questions, such as "What does X develop into over time?" and "Where does X move to over time?". While these questions are "locked-in" or "tethered" to a particular object other common causal-explanatory questions lack this feature.
These manufacturing and movement cases both meet a minimal interventionist criterion. In these examples each upstream change in the constitution or spatial location of an object "makes a difference" to an immediately downstream effect. 10 While 10 These examples may seem in tension with interventionism in the sense that cause and effect appear to reference the same object. Within the interventionist framework cause and effect must refer to sufficiently distinct properties such that it is possible to (hypothetically or conceptually) intervene on X without, at the same time, also intervening on Y. If this is not possible it suggests that, instead of X being a cause of Y, that X might just be Y (and vice versa). For example, consider the claim that turning a door handle to the clockwise position causes a "dextrorotatory" handle, or that introducing a disease gene into an organism causes a "mutated" genome, or that opening a refrigerator causes the door to be ajar. In each of these examples, instead of the purported cause producing the effect, both merely re-describe the same phenomenon. Examples of object-oriented causation do not have this issue because, although cause and effect are these cases primarily focus on changes to some single object of interest, they include many other causal factors that guide, channel, and direct the object as it undergoes these changes. In order to see this, consider manufacturing examples in which some starting material is used to produce a final product. These examples bear similarity to Aristotle's notion of "material causation" in which some material is the substance "out of which" a particular product is made (Physics B;194 b 26-9). 11 For example, a block of marble that is sculpted into a statue is a material cause of the statue. 12 The material cause is often associated with an "object" that is followed through the causal process as the marble "is not only the material out of which the statue is made; it is also the subject of change, that is, the thing that undergoes the change and results in a statue" (Falcon 2019). Of course, the starting material is just one cause of the downstream outcome, as other causes convert these materials into the final product. So, for example, just as a sculptor converts the bare marble into a statue, an enzyme converts a metabolite into a chemical product and developmental factors transform the egg into a butterfly. These "guiding" causes-sometimes referred to as "structuring causes" or "constraints"-are responsible for converting, channeling, or guiding the upstream starting material into a particular downstream outcome (Dretske 1988). Part of what this shows is that causes with material continuity are one of many causes in multicausal situations. 13 In these examples, we tend downplay the role of causes with material continuity, while emphasizing the role of guiding or structuring causes. This is evident in the fact that we are less likely to say that the building materials caused the house to be built, or that the block of marble caused the creation of the statue, or that the butterfly egg caused the butterfly to develop. This is associated with the view that starting materials alone do not produce the product as they remain motionless and inert unless acted upon by something else. As mentioned by Aristotle, "a statue does not come to be spontaneously" (Parts of animals; 640 a 30-35). 14 This de-emphasis on the causal role of starting materials is seen in the fact that they are characterized as a "passive cause," "passive matter," and as having "passive power" compared with the 11 (Lennox 2001a). 12 While my analysis of "material continuity" bears similarity to the notion of "material causation" discussed in the philosophical literature, they are not synonymous. The Aristotelian notion of material cause is highly complex and involves many considerations (necessity, goal-directedness) that are not part of this analysis of material continuity. Furthermore, while Aristotle's four causes can be understood as supporting different definitions of causation, the interventionist approach I rely on does not. This approach relies on a single definition of causation (interventionism) and distinguishes causes within this framework. 13 Notice again that this differs from the examples used to illustrate Woodward's three main distinctions, as in these cases we tend to identify a single main cause. 14 (Lennox 2001a, 640a, 30). associated with the same object, they refer to sufficiently distinct properties of the object. For example, in movement cases X and Y refer to different spatial locations, as there is interest in knowing whether the object's presence at one is causally relevant to its later presence at another. In manufacture cases X and Y refer to distinct forms-there is interest in knowing whether taking on one form X is causally relevant to the object then taking on another Y.
Footnote 10 (continued) 52 Page 10 of 17 more "active" role of guiding causes (Barnouw 2007;Noble 2017;Crombie 2020). 15 While we admit that starting materials are necessary for the outcome to occur, we do not always highlight their causal role in bringing about the outcome.
Why are starting materials often viewed as passive causes, while guiding factors are viewed as more active and causally responsible? There are at least two main reasons for this. First, while the presence of material is causally relevant to whether production takes place or not, it rarely determines which particular outcomes (of many potential outcomes) are produced. As the guiding cause is responsible for this it tends to be viewed as actively determining the outcome. For example, while the block of marble can be converted into a variety of different products (a column, statue, tile, etc.) the sculptor controls which is produced. In this case, causal responsibility and whether a cause is "active" is associated with the fact that the guiding cause has specific 2 causal control over the final product formed, while the starting materials do not. 16 A second reason why causes with material continuity are viewed as less causally responsible is that we often assume that they are present in a given situation, while other factors change in various ways. If we want to know how a house is built or how a butterfly a develops, we often assume that the starting materials are available.
While the explanatory role of material causes is often minimized, there are some exceptions to this. One main set of exceptions have to do with cases in which we assume that the guiding causes are fixed and unchanging. In many situations these guiding factors are viewed as "structuring" causes, not just because they channel the system to a particular outcome, but because they are fixed and less likely to change, relative to other causes. For example, given a set of water channels or blood vessels, we commonly assume that these structures are fixed and we focus on whether the material cause (the liquid) is introduced into this system or not. In this set up, availability of the object or starting material is what causes the process to unfold. 17 In biological contexts, organisms exploit this control of material causes-they sequester and release starting materials (such as glucose) as a way to stop and start biological processes (such as glycolysis), respectively. This is similar to having a ready sculptor but varying whether the marble starting material is available or not. When guiding causes are an assumed part of the background, or held fixed, causes with material continuity are viewed as more causally responsible.
Causes with material continuity are associated with a unique explanatory perspective that focuses on some material object and how it changes over time. These causes are more likely to arise in domains in which: there are material objects of interest, these objects have some stability within a relevant time-frame, and they undergo changes that are reliable, repeatable, and generalizable. As illustrated by the examples discussed in this section, causes with this feature arise in many scientific and ordinary life contexts.

Investigative strategy: tracer techniques
Causes with material continuity are often studied with a unique investigative strategy. These causes are often studied with physical tracer and tagging techniques, which exploit the flow of material through these causal processes. These techniques involve attaching a physical tag to the upstream cause and then following the tag as it moves through the steps of the causal process. Following this tag allows scientists to discover these downstream steps, where these steps can reveal different stages in an object's development, different locations it can travel to, and different ways that causal links are (and are not) connected up in some domain.
This basic approach is found in many scientific areas, including biochemistry, ecology, and neuroscience. In the context of biochemistry, scientists use tracers to identify metabolic pathways, which capture the sequence of chemical conversions that transform a metabolic substrate into a product. These cases involve attaching a radioactive tracer to a metabolite and following its journey through a sequence of chemical conversions. This reveals the "fate" of the original metabolite and the different chemical substances that it is transformed into on its way to becoming some final product. A second example comes from radioecology, in which ecologists use radioactive tracers to identify ecological pathways, which capture the prey-predator connections between species linked in food chains (Odum 1971). While it is often obvious which species are located in a particular area, it is much harder to determine the particular feeding relations among them. One way to identify this is to introduce a radioactive marker into an upstream prey (such as a species of grass) and then follow its movement into a sequence of downstream species at increasing time intervals. This allows ecologists to "unravel" and "untangle" these causal connections in ecosystems (Odum 1971). A final example involves the use of tracers to study neural pathways, which constrain the flow of nerve signals in the brain and other nervous tissue. In this context, neuroscientists use dyes, radioactive materials, and viruses to chart the flow of material along and across neurons, as this reveals the routes along which signals flow (Morecraft et al. 2014;Saleeba et al. 2019). Identifying this structural and anatomical connectivity provides insight into the function of different nervous tissues (Ross 2021b).
In scientific contexts, tracers are expected to meet particular criteria in order to provide reliable information about a causal system. 18 A first criterion is that the tracer should not alter the causal process being studied, as this would provide information about a perturbed system rather than the "natural" process of interest. Due to this, scientists will often claim that an ideal tracer should be "relatively harmless" and have minimal toxicity to the system (Hu et al. 2013;Huh et al. 2010;Mi et al. 2019). Second, tracers should be sufficiently recognizable in the causal process such that they can be properly identified and followed. The tracer should be distinguishable from surrounding materials and it should remain in high enough concentration in later steps of the causal process. Third, the tracer is often expected to be "specific" in the sense that it tags and marks exactly and only material that moves through the causal process (Martin and Dolivo 1983). One exception to this are tracers that reliably flow with material moving along a causal process, without the tracer binding to this material. 19 Tracer techniques are a unique causal investigative strategy in the sense that they are not used to establish or prove causality, but to get further information about systems that are already viewed as causal. This is supported by two main points. First, tracer methods require background information about the causal process, which is clear from tracer criteria. In order to use a tracer it needs to be clear what should be tagged, which requires knowing what (if anything) reliably flows through the causal process. Before using tracers in the biochemical, ecological, and neurological contexts above, scientists had already identified materials that reliably flowed through the causal systems. 20 This background causal knowledge was required for the tracer methods they employed. What scientists did not yet know is where the downstream steps of these pathways would lead and what would be found at each step. You can know that caloric energy flows through ecosystems and that cellular materials flow along neurons without knowing that the meadow frog eats the grasshopper and that the hypoglossal nerve innervates the tongue. Second, the fact that tracer methodology requires causal information is also suggested by the fact that movement of material alone is not sufficient for causation. We can identify and track the movement of some physical matter, without this outlining the steps of a causal process. For example, we can track the flow of material from an organisms's diet to the microconstituents of its cells, without this material being causally or explanatorily relevant to all of the cell's behaviors. 21 As tracking material alone is insufficient to provide reliable information about a causal system, something more is needed to ensure that tracer methods meet this standard.
A main suggestion of this analysis is that these tracer techniques can be used for some, but not all, causal systems. As these techniques involve tagging material and watching it flow through a causal process, they cannot be used for causal processes that lack this feature. Can tracer techniques be used to study causal systems that lack material continuity? If so, what makes this investigative strategy meaningful? To the extent that all causal relationships involve the flow of "causal influence" there is a loose (and less meaningful) sense in which "tracing" this influence can provide information about the system. A more compelling account of tracers for causal systems that lack material continuity would involve tracing "immaterial" entities such as the flow of "information" or "signals." There are a number of likely examples of these cases, including the use of genetic markers in biological systems and "traitor tracing" in social organizations. 22 While these signal or informational tracers are likely to share various features with tracers that tag material, they still have a number of important differences. Tagging material is a different technique than tagging information or signals-many systems will be better studied with one technique over another. This difference is likely associated with different explanatory perspectives-the focus on a persisting object in one case, and a persisting signal in another. Finally, to the extent that signal tracers appear less common than physical tracers, it is worth considering why this is the case. It may be that signal tracers are more difficult to implement for reasons having to do with complications in defining what counts as information, challenges in identifying this, and distinguishing information from causality.

Causal control: microstructure of effect
Material continuity differs from Woodward's causal distinctions in an interesting way. In particular, material continuity concerns the microstructure of cause and effect, while stability, specificity, and proportionality are silent on this aspect of causation. This can be appreciated by examining the type of "causal control" that each distinction among causation is associated with. For example, the control that a cause has over its effect can be stable, it can be fine-grained versus coarse-grained (specificity 1 ), or it can be specific in the sense of producing a single outcome, versus nonspecific in the sense of producing many outcomes (specificity 2 ). Importantly, causes with material continuity provide control over the microstructure of their effects. For example, while the sculptor controls which particular type of product is created-column, tile, or statue-the starting material controls its microstructure as it controls the physical material it is made of. Changing the starting materials provided to the this means is that, if some material is followed, the mere fact that it ends up in some system is no guarantee that it explains some feature of the system that we are interested in. sculptor-whether they are marble, granite, soap, or plastic-changes and controls the microstructure of the final product. This captured in an Aristotelian conception of material causation, as "[t]he material cause explains out of what kind of matter the effect comes," that is to say, what the effect is made of (Dodig-Crnkovic 2003, 531).
The fact that causes with material continuity have this type of causal control is important for a number of reasons. 23 First, microstructure matters because it is often explanatorily relevant to behaviors of the object of interest. Whether a final product is radioactive, magnetic, or buoyant depends on the material it is made up of, which is often determined by the starting materials used. In these cases, choice of starting materials will be explanatorily relevant to features of the final product. This helps clarify how properties of the final product are explained by the starting materials and not the craftsman-for example, that the marble statue will sink in a body of water is explained by the chosen starting materials, as opposed to the skill of the craftsman. Changing the starting materials (for example, to styrofoam) would change whether the statue exhibits this sinking property or not.
Second, in the context of manufacture cases, this microstructural control relates to the functionality of the final product. If the final product is supposed to serve some function-e.g. the feature of being waterproof-the chosen starting material will impact this and this will restrict which materials can be used. In this sense, "the nature of the craft product constraints the choice of appropriate materials" (Lennox 2001b, 289). This is due to the fact that, "whatever material is used has a nature of its own: the craftsman cannot do anything he likes with his material, but only what it is capable of being compelled to do" (Lennox 2001b, 289). This is relevant to creating products that are functional and it shows how the craftsman is limited by the starting materials as they provide causal control over the final product.
Third, this microstructural control provides a way to "import" material into the final product in ways that can be accidental or deliberate. Accidental cases include situations in which some toxic material in the cause delivers toxic material to the effect. For example, radioactive materials can enter into the constituents of food products, which are then incorporated into the organism's that ingest them. 24 While these cases involve accidental contamination, the same control can be used intentionally to move material to a particular location. This is exploited in some forms of drug delivery in which drugs are incorporated into materials that are then assimilated by particular areas of the body. If these drugs are toxic, they can destroy the specific cells that engulf them (which is valuable in cancer treatments). Some of 23 Consider other causes that appear to have microstructural control over their effects, but that lack material continuity. Heat can change the microstructure of water by converting it from liquid to gas. Chemicals can change the microstructure of metal by causing it to rust. While these causes exert control over an effect's microstructure, they can do so without exhibiting material continuity as they need not transfer material to the effect. Does this show that microstructural control is not unique to causes with material continuity? The important point here is that causes with material continuity exert control over the microstructure of some effect by transferring material to it-they "share" microstructure with their effect, which is not the case in the above examples. The block of marble shares microstructure with the statue in a way that heat does not share microstructure with water and that chemicals do not share microstructure with the metals they oxidize. 24 This happened after the nuclear disasters at Chernobyl and Fukushima Daiichi, for example. these delivery methods are said to implement a "Trojan horse" approach, as products that are normally welcomed into various cells of the body are loaded with various drugs, allowing them to be smuggled into the cell.
Finally, microstructural control can also be used to get information about a cause by studying its effect. The fact that cause and effect share microstructure means that there are physical "traces" of the cause in the effect such that studying the latter provides information about the former. For example, studying the materials that a statue is made of can identify the region in which it was created or where the materials were sourced. Additionally, radiocarbon dating can be used to date the time period of the cause and when it produced the effect, as is done with various archeological finds, including artifacts and fossils. Of course, other features of an effect, besides its microstructure, can reveal information about its cause. A sculptor might have a special carving technique that is used to identify their pieces. These features are likely to provide different types of information about the causal process-one about the craftsman and the other about their materials.

Conclusion
This paper has considered a novel distinction among types of causation, referred to as "material continuity," which concerns the reliable transfer of material from cause to effect. This distinction has been compared to commonly accepted causal divisions, including stability, specificity, and proportionality (Woodward 2010). While study of these divisions is commonly motivated by an interest in identifying causes that provide deeper and better explanations, this paper suggests another motivation-namely, capturing the diversity of causal structures in the world. Capturing this diversity is important, because it reveals how different types of causes involve unique explanatory perspectives, are studied with distinct causal investigative strategies, and provide different types of control over the world. Why exactly is capturing this causal diversity important? What turns on appreciating material continuity and the unique features associated with it?
We take interest in causation because it helps clarify how we navigate the world, provide scientific explanations, and control outcomes and bring about change. Appreciating the diverse causal structure of the world matters because different structures teach us different lessons for reaching these goals. Navigating causal systems that are unstable requires a different strategy set than systems that are stable. Explaining outcomes with non-specific, heterogenous causes involves various challenges that are not present when specific, single main causes are at play. The control available to us changes depending on whether causes have fine-grained specificity or are material causes with microstructural control. In each of these cases, the manner in which we select relevant causes, frame an explanatory-why question, provide an explanation, and gain control over the world will differ depending on the causal structure we are presented with. Appreciating distinctions among causation provides a fuller picture of the causal complexity of the world and the nuanced ways in which we study, navigate, and explain it.